Dear CASTEP users and developers,
I am using CALYPSO, a structure prediction software, which is interfaced with CASTEP. Randomly generated structures and evolved structures will be optimized by CASTEP.
I used 16 cores here, during structure prediction, there are errors like application called mpi_abort on all 16 cores, but the calculations haven't been interrupted until the job was done. When the calculations finish, there are castep.00nn.err files in outputs file list, as indicated in the attachments. Two different error reports are also attached.
My question is what those kinds of errors mean？Will the errors ,norm_sq is negative in wave_snormalise_slice_b, wave_orthonormalise_overlap_wv_ks: algor-invert failed in inversion, stop "current trace stack" ? Why I can still get a lot of optimized structures?
These errors almost certainly mean that the atoms were too close together in the initial configuration; so close, in fact, that the pseudopotential core regions overlap, which causes a breakdown of the ultrasoft pseudopotential formalism.
I recommend you look at the interatomic distances (e.g. with a CASTEP dryrun, or a visualiser) and move apart any atoms which are too close.
As a side-note, this kind of question would probably get more attention on the CASTEP email list:
https://www.jiscmail.ac.uk/cgi-bin/webadmin?A0=CASTEP
Hope that helps,
Phil Hasnip
(CASTEP Developer)
- Badges
- Science topic
- Similar topics
- Physical Sciences
- Materials Science
- Materials
- Inorganic Materials
- Metals
- Alloys